The development of the AAAQ was conducted systematically, and its psychometric properties were evaluated by testing it on a sample of 404 middle-aged persons (EFA and CFA only). The development of the AAAQ was guided by published literature and the Active Ageing Framework endorsed by the WHO , as well as input and feedback from experts in public health, geriatrics and gerontology. The AAAQ tool was intended to assess whether participants are aware of the three pillars of active ageing.
Most of the available tools have been developed for a specific population and set of objectives, which restricts usability for the general population, and which means that they are not appropriate for comparison. For example, the active ageing measurement tool developed by Perales et al. , targeted at European countries while  developed the tool to assess active ageing attributes based on Thai culture. Also, the evaluations of the related available tools to measure awareness of ageing have limitations in terms of their validity and reliability, and the issue being measured [18, 34].
Moreover, some of them have not completely incorporated all three pillars of active ageing, namely, health, participation, and security, into the questionnaire [17, 19]. Furthermore, the validation of these tools is not fully explained, and none of them have been translated and validated for local users. Therefore, the researcher became interested in developing an awareness of active ageing questionnaire for Malaysia based on the ageing framework selected for this study.
EFA is a statistical technique that is appropriate for scale development and is used to test or measure the underlying theory for hypothesized patterns of loading . During the development of the items, initially, three factors were developed to represent the three pillars of the Active Ageing Framework. However, it became apparent from the EFA result that the three pillars were inextricable. The EFA yielded two factors for the 22 items that explained 58 percent of the variance in the items at a significance level of less than 0.001, which indicated as good explanatory power. It was apparent from the EFA result that all the items appropriately belonged to these two respective factors, and these two factors were renamed health and non-health factors. The EFA also revealed that two items were highly correlated with each other, and it was noted that this issue might affect the model fits later.
CFA is a statistical technique that is used to verify the factor structure in a set of items. It allows the researcher to test the hypothesis that there is a relationship between the items and the factors to which they belong . Apart from a factor loading, convergent validity in this study was examined by observing the CR and AVE for each factor in the AAAQ. The value of CR and AVE should be more than 0.5 and 0.6, respectively. A lower CR indicates that the items do not measure what they are intended to measure. A low AVE indicates that more error remains in the items than the variance explained by the intended factor .
In the CFA, the initial 22 items were loaded into two factors (health and non-health). Then, a repeated process of modification was performed based on the factor loading of each item, the correlation between the items and factors, as well as the model fit. Initially, all the items with a factor loading of less than 0.4 were deleted. Also, when two items were highly correlated, the one with the lower factor loading was deleted. The MI were also examined, and when there were pairs of items that had high values, one of them was deleted or they were set as free parameters. Finally, the model that exhibited the most acceptable fitness based on RMSEA, GFI, CFI, TLI, and x2/df were kept. The findings suggested that the model with 14 items had the best fit, and achieved construct validity, which indicated that the AAAQ was able to distinguish between those who were aware of active ageing and those who did not. Thus, in the final model, eight out of the 22 items were deleted (36 percent). It has been suggested that the proportion of items deleted should not be more than 20 percent .
The deleted items were essential factors for active ageing as agreed by the experts, but the psychometric assessment among study population did not support this view. Most probably, the item statements were not well understood by the participants. Besides, the development of the items was based on the findings and opinions of older persons, which the participants, who were in a younger age group, may not agree with until they became elderly. The CR, AVE and standardised factor loading of the final model with 14 items indicated that convergent validity was achieved. The two structures in this scale were considered good because all the factor loadings of the items were more than 0.5, the AVEs were more than 0.5 and the CRs were more than 0.6. This result indicated that there was sufficient convergent validity in this tool. Thus, the items were well correlated with their respective factors. Cronbach’s alpha, AVE, and CR values confirmed that the AAAQ had internal consistency.
On the other hand, the AAAQ did not achieve discriminant validity. The maximum shared variance (MSV) and average shared variance (ASV) for both the health and non-health constructs were found to be 0.86. Thus, the values of both the MSV and ASV were higher than that of the AVE of the constructs, which suggested that discriminant validity was not achieved. However, even though discriminant validity is not achieved by the AAAQ, the overall validity of the AAAQ is not affected. In the AAAQ, all 14 items covering both constructs are summed up to get a score for the awareness of active ageing as there is no intention to discriminate the awareness of active ageing between the two constructs. This is because the review of the literature confirmed that the pillars of the Active Ageing Framework are inextricable. The fact that the three pillars of the Active Ageing Framework are inextricable may be the reason for the lack of discriminant validity [37–38].
This study has a number of limitations. Firstly, the result is based self-reported measure, which is a type of measurement that is prone to response and information bias. Secondly, the psychometric assessment of the AAAQ involved only the Malay version. It may be quite challenging to validate the English version in this study population as English is not their main language. Finally, as this is one of the first instruments to incorporate the Active Ageing Framework, there is no gold standards against which to evaluate its criterion validity. Thus, criterion validation of the AAAQ cannot be established.
In short, although the outcome of the EFA and CFA yielded only two factors, namely health and non-health factors; the essence of the active ageing framework is still captured in this questionnaire as it considers all three pillars, namely health, participation and security. Therefore, the 16 items (two stand-alone questions and 14 statements) of the Malay version of the AAAQ are satisfactory reliable and valid so the AAAQ can be used as a measurement tool to assess awareness of active ageing.